Difference-In-Difference

[Arthur Brahm]

Hello, I am trying the did method in panel data to see if an area that has built an 'economic infrastructure' has an impact on the area after it was built/intervention. There are 7 provinces in Indonesia with Regency/City units, therefore I also use the i.prov control

1. In the treatment variable/area that has the infrastructure, it is significantly positive towards local generated, grdp, poverty rate.
2. in the post variable/year after intervention in the area, it is also significantly positive towards local generated, grdp, poverty rate.
3. Here is my confusion, when Did = treatment*post, why is there nothing significant but only positive for lnlocalgen and even be negatif for lngrdp ?

Is this normal?

I use the command "reg treatment post did controlvariables i.prov" Thank you all

[Carlo Lazzaro]

Arthur:
if you want to go did with -regress- see PowerPoint Presentation (princeton.edu).
Otherwise, you can benefit from -didregress- or -xtdidregress- that allow you to test the pre-treatment parallel assumption.

[Jared Greathouse]

Okay, now I must shamelessly plug my new Forward DID command. It may not be as useful here, but it chooses the optimal combination of controls to do DID with. It's called fdid. It may install it like

net inst fdid, from("https://raw.githubusercontent.com/jgreathouse9/FDIDTutorial/main") replace
net get fdid, all